|
ETA on stereo processing
Andrew F. from United States [5 posts] |
17 year
|
Hello,
I just stumbled upon this project and have a couple questions. About how long will it be before we can get depth from stereo using Roborealm? (excuse me, I am just very excited about getting that). Also, I just want to verify that I can call the functionality of Roborealm from a C/C++ programmatically, and can send jpeg's or bitmaps to roborealm on my own, though a programmatic interface (and then read back the results of the roborealm image operations performed on the image).
Thanks in advanace,
Andrew F.
|
|
|
|
Stereo Vision
Anonymous |
17 year
|
Hi Andrew,
Thanks for the request. The answer is more about how you the users feel about certain features. The stereo vision request is rapidly climbing in terms of priority and the more you can tell us about how you want to use stereo the sooner we can schedule it. For example, do you have two images that you can share here? This would help us understand the intended use and quickly create some prototypes.
And yes, RoboRealm works very well as an IMPS (Image Processing Server) so you can send it images, have it process them and return back the results as an image or data variables using the API from any language that supports sockets. Nice huh? :-)
STeven.
|
|
|
|
Anonymous |
17 year
|
|
Stereo imaging is at the top of my list. I plan on using RoboRealm for a lot of things and being able to calculate distance would be a great feature to add. I see it as being a combination of taking two images (one would need to be able to specify how far apart the cameras are physically mounted) and comparing the two images to see how far one way or another the pixels are shifted. Using this method, distance can be calculated quite easily. Both cameras would need to be able to take shots at exactly the same time, exposure length, etc. otherwise, using it for navigation on a moving robot would be useless. Also, the ability to generate a real-time anaglyph would be incredible. One could then actually see the 3D world in front of the robot. Of course, the computing power needed for generating anaglyphs at 30fps might be a bit much. :)
|
|
|
|
some samples
Andrew F. from United States [5 posts] |
17 year
|
Here are some sample images. The images were taken without compression using 16mm lenses from two cameras separated by about 3 or 4 inches. Currently, the main goal of the depth perception is to do foreground detection. Our goal is to have our robot manipulate objects with it's grippers. We use line segment extraction and want to be able to tell which line segments belong to the item in the robots hands and which don't. We can paint his arms and grippers blue or green so that they are taken out. While we are still writing the object manipulation code, we are using a lazy susan, which operates similarly in principle (so for now, we are trying to pick out which line segments are part of the object sitting on the lazy susan).
We actually have the cameras mounted so that they can turn inwards so it would be incredibly useful for the algorithm to take an estimate of how "turned in" the cameras are (at first we are fine with parallel cameras, and that is how the sample images I'm sending were taken). We are interested only in relative distance at the moment. When we finally get this running on the robot (for now we are manipulating the objects by rotating them on a lazy susan a couple feet away). In our goal scenario, the objects are roughtly 6 inches away from the robots cameras, which are separated by about 3 inches.
Our lens is farely zoomed in at the moment, to approximate the resolution of the fovea (which is very roughly guestimated as a 5 degree field-of-view with a 200-pixel diameter, which is why we are using 16mm on a 1/3" sensor at the moment).
Thanks for your useful questions. It's helpful for me to get the parameters straight in my head as well.
-Andrew
|
|
|
|
Anaglyphs
Anonymous |
17 year
|
Re the anaglyphs: B&W anaglyphs arent hard to do in RR (see the attached Robo file) and the attached images  , .
-this pair was taken by moving the camera in my hands, not very precisely -so they arent 100% correct.
There some really nice (professionally done) pairs at http://www.pokescope.com/images/categories/Contests2006.html
I think it would be nice to have RR generate a single greyscale image that represented scene depths and then we might be able use thresholding etc to mask out near or far objects etc.
BTW I have noticed something a bit peculiar with 'Load Image' ;If I cut a single image into two halves (with Irfranview )and then load the two images, RR seems to cease working ? maybe its just a general wobbliness with load image rather than something to do with having cut them from one original image, though? [I was messing about with the wonderfull toad image from the site above.]
-Actually this toad has got to be a pretty good test -if RR can seperate him from the background (especially his front left leg) then it will be doing extremely well, I think!
Chas
 program.robo
|
|
|
|
Load image 'wobbliness'
Anonymous |
17 year
|
|
Hmmm - the load image module seems to be working fine ;so the only wobbly thing was me :-) Chas
|
|
|
|
Re. Manipulator
Anonymous |
17 year
|
Andrew, it looks as if youve got a lot of puzzles to solve to get your machine running, so I wonder if it might make life easier if you chose an initial object that was easy to filter out, like a green ball, you could then use the methods in the ball tracking tutorial and perhaps the new circle module to get X and Y values.
The idea of working with variable focal lengths and inward pointing cameras also seems terrifying and if your trig is as rotten as mine you might this tree height formula usefull:
http://www.nativetreesociety.org/measure/new_measurement_method.htm
In fact if your trig is as bad as mine, I would be tempted, initially, to stick the camera reasonably high with its focal plane parallel with the ground whilst you get the other bits going. - If you do work out the stereo maths please post it, as I for one would like to know what the answer is ! Thanks Chas
|
|
This Thread
